The Cyrix 5x86 is interesting for technology nerds (like me), because it has some pentium technology built into the core. For a user, it's a good processor for late 486 boards. If you enable the branch target buffer, it will get significantly more performance than an AMD 5x86, which is basically just a 486DX4 with a higher multiplier. On the other hand, I have yet to see an AMD 5x86-133 that doesn't overclock without any issues to 160MHz, but many Cyrix 5x86-100 already struggle at 120MHz, especially if you enable branch brediction. They say IBM branded Cyrix 5x86 processors overclock slightly better than Cyrix branded ones, so getting that system working at 120MHz is not unlikely (but not guaranteed).
If your primary goal is just playing retro games on period correct hardware, I wouldn't recommend a Cyrix 5x86 system over an AMD 5x86 system. On the other hand, AMD 5x86 systems are boring and everyone and their dog already have one, so if you want to have "something special" or your goal is to tinker with special setups (like running at FSB60), you might want that system to get the parts. The Cyrix 5x86 is interesting for FSB60, because the 100MHz version supports a *2 multiplier, allowing 2*60 = 120MHz, which works on many 5x86-100. On the other hand, the AMD 5x86 only supports *3 and *4, so the lowest core clock at FSB60 is 180MHz, which only works on selected processors. On the other hand, when trying to run at FSB60, you will quickly find out that most 486 boards are not designed to operate at that speed, and you likely need to add that many wait states that the higher FSB clock doesn't help at all. Finally, if you want FSB60, you could also look into getting an AMD DX4-120, which is fully in specification at 120MHz and supports a *2 multiplier. (EDIT: Just to be clear: I didn't intend to claim that 2*60 is "fully in spec". The FSB clock of the DX4-120 is specified up to 40MHz, or perhaps 50MHz at 2*50 = 100MHz, but is is definitely out of spec at 60MHz.) This processor is less finicky at 120MHz, but provides less performance. My coarse experience with some benches from Phils DOSBENCH suite is that a Cx5x86 with branch prediction enabled is approximately 33% faster than an AMD 5x86 at the same clock.
